communication in nano-scale via fretbrus/ic1004-td-14-09073.pdf · scale components. they are...
TRANSCRIPT
EUROPEAN COOPERATION
IN THE FIELD OF SCIENTIFIC
AND TECHNICAL RESEARCH
—————————————————
EURO-COST
—————————————————
IC1004 TD(14) 09073
Ferrara, Italy
5-7 February, 2014
SOURCE: Jagiellonian University,
POLAND
Department of Telecommunications,
AGH University of Science and Technology,
POLAND
Communication in nano-scale via FRET
Krzysztof Wójcik, Kamil Solarczyk Jagiellonian University
Ul. Gołębia 24
31-007 Krakow
POLAND
Email: [email protected], [email protected]
Pawel Kulakowski
Department of Telecommunications
AGH University of Science and Technology
Al. Mickiewicza 30
30-059 Krakow
POLAND
Phone: +48 12 617 3967
Email: [email protected]
Communication in nano-scale via FRET
Krzysztof Wojcik, Jagiellonian University
Kamil Solarczyk, Jagiellonian University
Pawel Kulakowski, AGH University of Science and Technology
ABSTRACT Nano-communication has gained significant attention in the last few years, as a means to
establish information transfer between future nano-machines. Comparing with other
communication techniques for nano-scale (calcium ions signalling, molecular or catalytic
nanomotors, pheromones propagation, bacteria-based communication), the phenomenon
called FRET (Förster Resonance Energy Transfer) offers significantly smaller propagation
delays and high channel throughput. In this paper, we report our recent experiments on FRET-
based nano-networks performed in the Laboratory of Cell Biophysics of the Jagiellonian
University, Kraków. We propose to use Alexa Fluor dyes as nano transmitters and receivers,
as they enable to create MIMO-FRET communication channels and thus enhance the FRET
efficiency. We measure FRET efficiency values, calculate the BER for the measured
scenarios and extend the calculations to consider a general case of MIMO (n,m) FRET
channels.
I. INTRODUCTION
Nano-machines are envisioned to be artificial mechanical devices that rely on nanometer
scale components. They are predicted to be basic building blocks for future nano-robots and
nano-processors [Akyildiz-2008, Suda-2005]. Their existence will create the possibility to
manipulate the animate and inanimate matter on a very low level. They promise an enormous
spectrum of applications: from biomedical (genetic engineering, health monitoring, medicines
delivery) and industrial (material fabrication) to environmental ones (pollution removing,
control over nature ecosystems). Because of its low complexity and size, the possible tasks
taken by nano-machines are very limited. In order to perform more complicated operations,
these devices must cooperate together, working as large well-organised structures, i.e.
constituting networks. Hence, the crucial issue is to enable efficient communication between
nano-devices.
The current progress in the area of nano-technology spurred the development of its younger
sister: nano-communications. A large number of nano- and micro-scale mechanisms was
recently analysed from the communication view-point: calcium ions signalling [Nakano-
2005], flagellated bacteria carrying DNA-coded information [Gregori-2010], molecular
motors [Moore-2006, Serreli-2007], pheromones and pollen propagation [Parcerisa-2009] and
others [Nakano-2013]. Most of them, however, guarantee neither a good throughput nor an
acceptably low transmission delay.
Here, we consider the phenomenon of Förster Resonance Energy Transfer (FRET) as a means
for efficient nano-communication. It is a spectroscopic process in which one molecule
transfers its excitation energy non-radiatively over a distance of 1-10 nm to another molecule
[Forster-1948]. The FRET phenomenon has some critical communication advantages. The
first one is a very short timescale of the energy transfer that occurs in nanoseconds. Second,
the FRET probability depends on the matching between the emission and absorption spectra
of the molecules and decreases with the sixth power of the distance between them.
Consequently, FRET can be simultaneously used in different parts of a nano-network, like a
wireless spectrum band in a cellular network.
Nano-communication via FRET was already discussed in [Kuscu-2012, Kuscu-2013] where
information theory analysis and simulation results for FRET channels were given. In this
paper, we present experimental results of FRET-based communication in nano-networks of
antibodies (Immunoglobulin G) labelled with Alexa Fluor (AF) fluorescent dyes. Antibodies
have the ability to recognise some parts of specific molecules and thus they can be thought of
as nano-machines (more specifically: nano-sensors), while AF dyes act as their nano-
transceivers transmitting and receiving information bits encoded in quanta of excitation
energy. We propose to use AF dyes, as they enable multiple-input multiple-output (MIMO)
FRET communication. We give the measured FRET efficiencies, calculate bit error rates for
the measured scenarios and extend the calculations to consider a general case of MIMO (n,m)
FRET channels. The main contribution of this paper can be listed as following:
• a FRET nano-scale MIMO communication technique based on AF dyes is proposed,
• experimental results of FRET efficiencies for AF-based FRET networks are shown,
• bit error rates for MIMO (n,m) FRET channels are calculated.
The rest of the paper contains the following sections. In Section II, the FRET theory and the
feasibility of FRET multi-hop communication are considered. In Section III, the details are
given on constructing the AF-based FRET networks. Next two sections contain the
description of the laboratory methodology and the obtained measurement results. The bit error
rate calculations for MIMO (n,m) FRET channels are presented and illustrated in Section VI
and finally the paper is concluded in Section VII.
II. FÖRSTER RESONANCE ENERGY TRANSFER
Förster Resonance Energy Transfer1 (FRET) is an energy process involving two molecules.
The first one, called donor, should be in the excited state, e.g. after absorbing a photon. The
second molecule is called acceptor and should be in the ground state. The energy is
transferred from the donor to the acceptor (usually 1-10 nm) non-radiatively, without an
emission of any photon, by the long range dipole-dipole interaction of both molecules (the
explanation of FRET assumes that both molecules are dipoles oscillating with similar
frequencies that can exchange energy with each other [Lakowicz-2006]). The rate of the
energy transfer and the range of distances over which it can be observed depend mainly on the
overlap of the donor emission spectrum and the acceptor absorption spectrum for a given
donor-acceptor pair. These relationships were first measured by Förster [Forster-1948], and
later studied by Stryer and colleagues [Stryer-1978].
If the donor molecule is excited not by an external photon, but the energy comes from a
chemical reaction, the phenomenon is called Bioluminescence Resonance Energy Transfer
(BRET). BRET has some advantages over FRET. First, the excitation of the donor
1 The abbreviation FRET is sometimes resolved as Fluorescence Resonance Energy Transfer, which however
seems to be a misconception, as the phenomenon does require neither a donor nor an acceptor to be fluorescent.
Because of that reason, some authors just use the name Resonance Energy Transfer abbreviated as RET. Here,
we follow the most common norm, where 'F' in FRET stands for Förster, honouring Theodor Förster (1910-
1974), the author of the RET theory.
fluorophore in FRET may also lead to the simultaneous excitation of a small part of acceptor
fluorophores, which is obviously not desirable. Second, this excitatory light causes
photobleaching2 of the donor. Third, BRET signal to noise ratio is higher than FRET, thus
fewer molecules are needed to reach the same signal level.
II.A. FRET rate and efficiency
Let’s assume first that an excited molecule exists, but there is no other molecule (potential
acceptor) nearby. Then, the energy of the molecule is usually released by a photon emission.
It can be also dissipated by other non-radiative mechanisms like quenching or decay processes
(Fig. 1).
Fig. 1. The simplified Jablonski diagram showing possible energy transitions in an excited molecule.
The time a molecule spends in the excited state before returning to its ground state is called
the lifetime of the excited state of the molecule. The average lifetime of the molecule in the
absence of FRET transfer can be calculated as the inverse of the sum of its emission rate eµ
and its rate of non-radiative decay mechanisms (i.e. quenching or non-radiative relaxation)
nrµ :
nre
FRETno µµτ
+=−
1. (1)
The emission and the non-radiative decay are random processes with exponential
distributions.
It should be noted that some energy is usually lost in the molecule between excitation and
emission due to rapid decays between vibrational energy levels (see Fig. 1). The resulting
difference between the molecule absorption and emission spectra is called the Stokes shift3.
Now, if another molecule, being in the ground state, exists in the vicinity of the excited one,
the excitation energy can be also released by FRET. In order to have it happen, two conditions
must be met. First, both molecules must be in the distance of 1-10 nm from each other.
Second, the emission spectrum of the excited molecule, called donor, must overlap the
2 Photobleaching is a phenomenon of chemical modification of the fluorophore, which in turn loses its ability to
emit photons. This is usually caused by breaking open or bonding of the fluorophore to nearby molecules.
Although photobleaching is generally assumed to be irreversible, in some cases the fluorophore can be switched
on again after loss of emission ability.
3 For Sir George Gabriel Stokes (1819-1903), who was the first to observe this phenomenon in 1852.
absorption spectrum of the ground state molecule, called acceptor. The FRET rate can be
calculated as [Lakowicz-2006]:
∫∞
⋅⋅⋅⋅⋅⋅
⋅⋅
⋅=
0
4
456
2
)()(128
10ln9000λλλελ
πκµ
µ ddFnNr
AD
e
FRET (2)
The term eµ is for the donor emission rate and r is the distance between the molecules. The
factor 2κ is responsible for the relative orientation between the donor and acceptor dipoles
and is usually assumed to be 32 which is suitable when both molecules can move by
rotational diffusion [Lakowicz-2006], which is also the case of the experiments reported here.
The term N is the Avogadro number, n is the refractive index assumed to be equal to 1.4 for
the biomolecules in aqueous solutions. Finally the integral part of the equation is responsible
for the molecules spectra overlap: )(λDF is the normalised emission spectrum of the donor,
)(λε A is the absorption spectrum of the acceptor. As the FRET rate depends on the sixth
power of r , the phenomenon is commonly used in spectroscopy for very accurate
measurements of distances between molecules.
With an acceptor nearby, FRET is another possibility for the donor molecule to dissipate its
energy. In general, there can be more that one neighbouring molecule in the vicinity of the
donor and each of them can be the acceptor in the FRET transfer. All these neighbouring
molecules contribute to the total FRET rate. Thus, comparing with (1), the donor average
lifetime decreases and is equal to:
∑=
++=
k
iiFRETnre
FRET
1,
1
µµµτ , (3)
where iFRET ,µ is the FRET rate of the th−i possible acceptor among k ones. Knowing eµ ,
nrµ and all iFRET ,µ rates, we can also calculate the FRET efficiency iE , i.e. the fraction of
the excited donors that decay through FRET to the th−i acceptor [Lakowicz-2006]:
∑=
++=
k
iiFRETnre
iFRET
iE
1,
,
µµµ
µ. (4)
The FRET efficiency iE is also the probability that the energy is transferred to the th−i
acceptor. Thus, the total probability that the energy is transferred to any of the acceptors is
given by:
∑
∑
=
=
++=
k
iiFRETnre
k
iiFRET
E
1,
1,
µµµ
µ. (5)
The distance between the donor and the acceptor where nreFRET µµµ += is called the Förster
distance 0R , which can be measured or calculated with (2), for each donor-acceptor pair.
Knowing 0R , the FRET rate can be obtained from a simplified formula:
60 )()(r
RnreFRET ⋅+= µµµ . (6)
If there is only one possible acceptor (or others have negligibly small FRET rates), the FRET
efficiency can be calculated as:
60
6
60
Rr
RE
+= , (7)
what means that for 0Rr = , %50=E and FRETnoFRET −⋅= ττ 21 .
If there are many acceptors, the FRET efficiency is given by:
∑
∑
=
=
⋅+
⋅
=k
i i
k
i i
rR
rR
E
16
60
16
60
11
1
, (8)
where ir is the distance from the donor to the th−i acceptor. If the acceptors are equally
distant from the donor, (8) simplifies to:
60
6
60
Rkr
RkE
⋅+
⋅= . (9)
In practice, neither FRET rate nor its efficiency could be measured directly. FRET used to be
measured by observing changes in donor's and acceptor's fluorescence intensities. These
intensity-based methods, however, have a major drawback due to the influence of
photobleaching on measured FRET efficiencies. The alternative, more reliable method of
measuring FRET is based on fluorescence lifetimes. What can be observed in this kind of
measurements is the decrease of the average lifetime of the donor molecule after the acceptor
is added. Having FRETno−τ and FRETτ values and applying (1) and (3) to (4), we obtain:
FRETno
FRET
FRET
FRETnoFRETE−
−
−−
−
−=−
=ττ
τ
ττ1
1
11
. (10)
II.B. FRET time scale and throughput
Comparing with different nano-scale communication mechanisms, FRET occurs much faster.
The whole process can be divided into three phases. First, a donor molecule absorbs a photon
which takes less than a femtosecond ( 1510− s). Then, the excess of energy is dissipated rapidly
by the internal conversion to the lowest vibrational energy level 1S , which occurs in about a
picosecond ( 1210− s). After that, the FRET occurs after a random delay equal to the donor
excited state lifetime which is of order of nanoseconds. Finally, the donor molecule is again in
the ground state, while the acceptor is excited. Summing up the three phases, the delay of the
whole energy transfer is of order of nanoseconds. Assuming that a single information bit is
going to be sent each time (e.g. '1': the donor is excited, '0': no excitation), the FRET channel
throughput can be estimated as tens of Mbits/s.
II.C. Multi-hop FRET communication
The FRET phenomenon can occur between any types of spectrally matched molecules, see
(2). They can be of biological origin, i.e. organic molecules, or artificial ones, e.g. quantum
dots. In biophysics, it is very common to observe FRET for fluorophores, i.e. organic
molecules which emission spectra cover the range of visible light. For the purpose of
communication via FRET, the molecules emission/absorption spectra can be also located in
other parts of the electromagnetic spectrum, e.g. ultraviolet or infrared bands.
It is worth to note that the FRET phenomenon can be used for multi-hop transmissions in
molecule chains, where two of more energy transfers occur, one by one. In such a scenario,
the acceptor in the first FRET pair is at the same time the donor in the second pair, allowing
the signal to pass two or more hops (see Fig. 2a). Because of the Stokes shift, some energy is
wasted in each hop and the absorption spectrum for the second acceptor lies in the longer
wavelengths (lower frequencies) range that for the first acceptor (Fig. 2b). It means that
FRET channels are in general one-directional4: in Fig. 2a, the transfer A→ B → C is feasible,
but C → B → A is very unlikely. As the FRET rate decreases with the sixth power of the
distance, it is generally more efficient to transfer a signal via multi-hop FRET than directly.
One must assure, however, that all the FRET efficiencies are high enough. The total transfer
efficiency is the product of the FRET efficiencies for all the hops. FRET efficiencies in a
multi-hop transfer are independent of each other. It is confirmed by the Kasha's rule stating
that the emission spectrum of a molecule does not depend on its excitation wavelength
[Kasha-1950].
Fig. 2. a. An example of a chain of molecules where multiple FRET reactions can occur.
b. Respective emission and absorption spectra of A, B and C molecules.
III. FRET NETWORKS
The current level of nanotechnology does not allow us to create networks of nano-machines
freely moving, approaching each other and communicating. Instead, with the aid of
biotechnology techniques, we can construct structures made of proteins, potential nano-
machines, and attach to them some fluorophores which can communicate via FRET with each
other. It means that we cannot create any network structure, but we are rather limited by the
geometry of the available proteins.
4 There are some exceptions, see the homotransfer phenomenon which occurs e.g. between Bodipy fluorophores
[Lakowicz-2006] or YO-PRO-1 molecules [Hannestad-2008].
For the purpose of our studies, we built a FRET network composed of 3 antibodies (we used
Immunoglobulin G) labelled with fluorescent Alexa Fluor dyes. Antibodies are proteins that
have the ability to recognise some parts of specific molecules, thus they can be treated as
nano-sensors. The Alexa Fluor (AF) dyes are a family of fluorophores characterised by a good
robustness against photobleaching. Each fluorophore in the family is usually called Alexa
Fluor X, where X is the main wavelength, in nanometers, of its absorption spectrum. The AF
dyes can be easily labelled (attached) to antibodies, i.e. proteins of ca. 10-15 nm large. Thus,
fluorescent AF dyes can be considered as nano-FRET-transceivers attached to antibodies
working as nano-machines (nano-sensors).
While there are other components that can be used in FRET networking (see the summary at
the end of this section), we decided to choose the Alexa Fluor dyes mainly for their labelling
properties. Using these molecules, each antibody can be labelled not only with a single AF
dye, but with a large number of them (theoretically up to 10 dyes), depending on the degree of
labelling (DoL). In the result, we had a kind of multi-input multiple-output (MIMO) FRET
links between the antibodies. As explained in Section VI, having many AF dyes at both sides
of each link greatly increased the efficiency of the information transfer.
Our FRET network under study was constructed as follows. The first element of the network
was a histone H1 bound to a DNA molecule. Next, a primary antibody Ab0 was attached to
the histone H1. Then, Ab1, Ab2 and Ab3 antibodies were subsequently added (Ab2 and Ab3
were together attached to Ab1) in order to form the whole FRET network. Each of these 3
antibodies had been already labelled with AF dyes (see Fig. 3).
In general, antibodies allow building practically infinite chains, but the distances between
them cannot be controlled precisely. Our experiments were performed on HeLa 21-4 cells (a
common laboratory cell line derived from a human ovarian cancer) containing about 107 of
these histone H1 molecules in their nuclei. In the laboratory process of immunostaining, about
30–50% of histones are labelled with antibodies, thus we were able to create about 61053 ⋅−
similar fluorophore networks, which we assumed was a good statistical sample.
The main aim of our experiments was to measure the efficiency of the information transfer in
all the three communication links created between the antibodies: Ab1-Ab2, Ab2-Ab3 and
Ab1-Ab3. Each used antibody (Immunoglobulin G) had a shape of a 3-element airscrew [Tan-
2008] with a radius of ca. 7 nm (see Fig. 3). The distances between the antibodies Ab1-Ab2
and Ab1-Ab3 were about 9-12 nm, while Ab2-Ab3 distance is smaller, about 7-8 nm. The
antibodies Ab1, Ab2 and Ab3 were labelled with Alexa Fluor 405, 488 and 546 dyes,
respectively. AF dyes are rather small molecules, about 1.4-1.8 nm large. We used AF dyes
produced by Molecular Probes, Inc., having the following degrees of labelling:
• for AF 405, DoL = 2,
• for AF 488, DoL = 7,
• for AF 546, DoL = 4.
Due to the chemical properties of antibodies, the AF dyes always attach to their upper
elements. The exact position of the dyes on antibodies is, however, hard to determine. The
upper element of each antibody is about 7 nm long. Moreover, the dyes surround these
elements. As the result, the distances between the AF dyes may vary significantly. We can
estimate them as:
• AF 405 – AF 488 distances = 7-14 nm,
• AF 488 – AF 546 distances = 6-12 nm,
• AF 405 – AF 546 distances = 7-14 nm.
Fig. 3. The network of antibodies (Immunoglobulin G) and AF dyes used in the experiments.
The absorption and emission spectra of the AF dyes are shown in Fig. 4. The Förster distance
0R between AF488 and AF546 is equal to 6.4 nm [Lifetechnologies-1]; the 0R values for
AF405→AF488 and AF405→AF546 FRET transfers are, to the best of authors’ knowledge,
not published. The energy transfers in the opposite directions can be neglected as highly
unlikely, what can be confirmed with Fig. 4, as the respective emission and absorption spectra
do not match. It means that all the three communication links are unidirectional.
Fig. 4. Absorption (dashed lines) and emission (shaded areas) spectra of Alexa Fluor dyes [Lifetechnologies-2].
IV. LABORATORY METHODOLOGY
Since it is impossible to record the time decay profile of the signal from a single excitation-
emission cycle, fluorescence lifetimes were measured in the Time Correlated Single Photon
Counting (TCSPC) mode. This method is based on an accurate measurement of the time
between the excitation pulse and the arrival of the first photon to the SPAD (Single Photon
Avalanche Diode) detector. Numerous repetitions of this measurement permit to build a
histogram, which represents the time decay of the signal. Fitting the resulting histogram with
an exponential curve allows extracting the lifetime of a fluorophore.
FLIM (Fluorescence Lifetime Imaging Microscopy) measurements were done using a Leica
TCS SP5 II SMD confocal system integrated with FCS/FLIM TCSPC module (Fig. 5). This
module consists of picosecond laser diode heads (excitation at 405, 470 or 640 nm), coupled
to a laser driver (PDL 828 “Sepia II”), two SPAD detectors (PDM Series), a TCSPC module
(PicoHarp 300) and a detector router (PHR 800). In a typical measurement, a FLIM image
was acquired at 256x256 pixel resolution and at a speed of 200 lines/s. The acquisition time
for each image was one minute, with the laser power set to achieve a photon counting rate of
200-300 kCounts/s (the detection efficiency was about 1-10%). The fluorophores, AF405 and
AF488, were excited at a 20 MHz pulse repetition rate with 405 and 470 nm laser lines and
the emission was collected with 460-500 and 500-550 band-pass filters, respectively.
It is worth to notice that during each laser pulse, not more than few donor AF dyes were
excited among millions of them in our HeLa cells. Consequently, we could neglect the
situations with 2 or more excited dyes attached to the same antibody, what is important for the
later considerations on MIMO-FRET efficiency (see Section VI).
Fig. 5. Schematic representation of the Laser Scanning Microscope (LSM) integrated with a FLIM module. The
experiment begins when the sample is excited with a pulsed laser, which is controlled by a laser driver. This unit
permits to control the laser power output and its repetition rate. Upon excitation, the sample emits fluorescence
photons which are detected in SPAD detectors. In order to determine the exact time between the excitation pulse
and arrival of the first photon to the detector, a TCSPC module connected to a computer receives information
from the detectors, laser driver and laser scanner. In steady-state measurements (image acquisition without
information about fluorescence lifetimes), the sample is excited with a CW (continuous wave) laser, signal is
detected in PMT detectors and the resulting image is viewed on a LSM computer.
In each measurement a region of approximately 50x50 µm containing 2-4 cells was chosen. In
a given region three lifetime images were collected for AF405: one before and after each
staining with secondary antibodies Ab2 and Ab3 (tagged with AF488 and AF546
respectively), while for AF488 two lifetime images were collected: one before and one after
staining with AF546-tagged secondary antibody. Before analysis, to exclude the signal
resulting from non-specific staining, only the nuclei of cells were chosen. Due to possible
differences in the distance between attached antibodies resulting from chromatin compaction,
mitotic chromosomes were not analysed (see Fig. 6 A-F). The analysis was done in
SymPhoTime II software (PicoQuant GmbH) using tail fitting, excluding the IRF (Instrument
Response Function) time range. To improve signal to noise ratio, all nuclei in a given region
were analysed together, resulting in one decay for each image. Both AF405 and AF488
decays were fitted with a two-exponential function:
21 /2
/1)( ττ tt eAeAtF −− ⋅+⋅= (11)
where τ1 and τ2 are the lifetimes, while A1 and A2 represent the amplitudes of the components.
The goodness-of-fit was estimated based on the weighted residuals and the chi squared value.
For FRET efficiency calculations, see (10), the amplitude-weighted average lifetime was
used:
∑
∑ ⋅=
i
i
i
ii
A
A ττ (12)
The images of the observed cells (see Fig. 6) were obtained with a Leica TCS SP5 II SMD
confocal system (Leica Microsystems GmbH), equipped with a 63x NA 1.4 oil immersion
lens. Images (8 bit) were acquired at 512x512 pixel resolution, at a speed of 200 lines/s, with
1-3 frames averaged. For AF405 excitation, a 405 nm pulsed laser line was used, while Alexa
Fluor 488 and 546 were excited with 480 and 543 nm lines (Ar and HeNe lasers).
Fluorescence emission was collected using photomultipliers at 460-500, 500-550 and 580-630
nm, respectively. For the comparison of images before and after staining with secondary
antibodies, the same laser power and photomultiplier gain was used.
V. MEASUREMENT RESULTS
As explained in Section II.A, the FRET efficiency cannot be observed directly, i.e. measuring
the FRET rate. Instead, the lifetime of the excited state of the donor molecules is compared in
two scenarios: without and with the acceptor molecules, see (10).
It should be emphasized that, unlike the FRET efficiency, fluorescence life time of a dye
depends on pH, ion concentration and presence of other molecules in the solution. Thus, both
lifetime measurements, without and with the acceptor molecules, should be performed in the
same environment. Having this in mind, we always measured both fluorescent lifetimes on
the same selected population of molecules: there were 2-4 cell nuclei (see Fig. 6), each
contained ca. 3-5 x 106
FRET networks.
FLIM (Fluorescence Lifetime Imaging Microscopy) measurements showed the following
FRET efficiencies:
donor acceptor FRETno−τ [ns] FRETτ [ns] E [%]
AF 405 AF 488 4.91 4.20 14 AF 488 AF 546 3.26 1.75 46 AF 405 AF 546 3.08 2.79 9
Tab. 1. The results of the performed FLIM measurements
Additionally, FRET can be observed visually on confocal images by the decrease of the donor
fluorescence intensity in the presence of the acceptor, as shown in Fig. 6.
Fig. 6. The decrease of the fluorescence signal as a result of FRET transfers: AF 405 → AF 488, AF 488 → AF
546 and AF 405 → AF 546. HeLa cells were fixed and incubated with a primary antibody against histone H1.
Subsequently, the cells were stained with secondary antibodies (AF 405, 488 and 546) and images were
collected before and after staining with each antibody. A) A405 fluorescence before staining with A488. B)
A405 fluorescence after staining with A488. C) A488 fluorescence before staining with A546. D) A488
fluorescence after staining with A546. E) A546 fluorescence in cells previously stained with A405 and A488. F)
Transmitted light image of the cells showed in A-E. G) A405 fluorescence before staining with A405. H) A405
fluorescence after staining with A546. I) A546 fluorescence in cells previously stained with A405. J)
Transmitted light image of the cells showed in G-I.
FLIM measurements and confocal images confirm that the FRET transfer occurred in all three
communication links. As we do not know the exact distances between the AF dyes, it is hard
to verify if the measurement results match the FRET theory. Anyway, we can try to do that
for the FRET AF488→AF546, as we know the respective Förster distance: nm 4.60 =F . In
this link, for each donor we have 4 acceptors (for AF 546, DoL = 4). Assuming that each of
these 4 acceptors is located at the same distance r from the donor, we can put the measured
FRET efficiency and the Förster distance into the equation (9) and calculate: nm 3.8=r . The
calculated value matches our expectations: in Section III, the average distance between AF
488 and AF 546 dyes was estimated as 6-12 nm.
VI. MIMO-FRET CHANNELS
The FRET efficiency measured for all three communication links seems far from being
acceptable for telecommunication purposes. Let us adopt, like in [Kuscu-2012], the simple
ON-OFF keying modulation scheme. We can transmit a single bit '1' – exciting the donor and
hoping the energy is transferred via FRET to the acceptor – or a bit '0' – keeping the donor in
the ground state, i.e. not exciting it. Assuming no other source of excitation and no other
donors in the vicinity, transmission of '0' is always successful and transmission of '1' is
erroneous with the probability E−1 . If zeros and ones are equally probable to be transmitted,
the channel bit error rate (BER) is equal to ( )E−⋅ 15.0 , what means that even in the best
measured case (AF488→AF546) the BER is very high: 27%.
Obviously, the channel BER can be decreased by repeating each information bit numerous
times, at the cost of the channel capacity. Another possibility is to exploit multiple AF dyes at
both sides of each communication link. As stated in Section IV.A, in the performed
experiments we had only one donor AF dye excited at the same time. If more donor dyes
(attached to the same antibody) are excited, there is larger probability that at least one of them
transfers the excitation energy via FRET to the donor side. Thus, we can transmit '1' by
exciting as many donor dyes as possible. Then, the correct reception takes place when at least
one acceptor dye receives the energy via FRET. Transmission of '0' is realised as previously,
i.e. not exiting any donor. By using multiple dyes at both sides of the link we fully exploit the
multiple-input multiple-output (MIMO) character of the communication channel between the
AF dyes.
Let us assume that the FRET efficiency E in the ( )m,1 communication channel is equal
independently which donor AF dye is excited. In fact, what we have measured is the FRET
efficiency for different donor AF dyes averaged over tens of millions of excitation events5.
We can now give the probability of the correct reception of a bit '1' in the ( )mn,
communication channel which is the probability that at least one donor AF dye transfers its
excitation energy to an acceptor dye:
( )nEP −−= 11 . (13)
Consequently, the bit error rate for the ( )mn, channel is equal to:
( )nmn E−⋅= 15.0BER , , (14)
5 The photon counting rate at the detector was at least 200 kCounts/s, each experiment lasted not less than 1
minute and the maximum detector efficiency was 10%, thus we had at least 120x106 photons emitted. The FRET
efficiency was at least 6% what means we had at least 127x106 excitation events.
where E is the FRET efficiency for ( )m,1 communication channel measured experimentally
or calculated with (5) or (8). Therefore, we can calculate the mn,BER values for all three
communication links. These are the expected bit error rates in the case when all donor AF
dyes are excited each time a bit '1' is to be sent through the communication channel. The
mn,BER values are given in Table 2, together with measured FRET efficiencies.
donor donor
DoL acceptor
acceptor
DoL
MIMO channel
dimensions
E for (1,m)
channel
BER for
(n,m) channel
AF 405 2 AF 488 7 (2,7) 14 % 37 % AF 488 7 AF 546 4 (7,4) 46 % 0.7 % AF 405 2 AF 546 4 (2,4) 9 % 41 %
Tab. 2. Summary of measured FRET efficiencies and calculated BER values
We can see that the link AF488→AF546 is clearly the best and the only one with the
satisfactory BER value. On the one hand, it is the result of a bit smaller separation between
the donor and acceptor molecules: they are on average 1-2 nm closer than in other two links.
Such a difference can be crucial, as the FRET rate decreases with the sixth power of the
donor-acceptor separation, see again (2). On the other hand, it is the large number of Alexa
Fluor dyes at both donor and acceptor sides (11 in total) that enables to reduce the channel
BER. As a reference, we also calculated mn,BER values for some theoretical cases: donor and
acceptor DoL varying from 1 to 10 and donor-acceptor separation from 05.0 F to 07.1 F (see
Fig. 7). We can see that BER values are increasing rapidly with donor-acceptor distances. The
acceptable BER (about 1%) can be obtained even for large distances ( 05.1 Fr > ), but at the
cost of huge number of AF dyes involved. It also seems better to have balanced distribution of
dyes at both sides of the link, or a slightly more dyes at donor side than at acceptor side
(compare e.g. (4,4), (2,6) and (6,2) cases).
Fig. 7. Theoretical BER values for (n,m) MIMO-FRET channels
VII. CONCLUSIONS
In this paper, we discuss MIMO communication in nano-networks realised via FRET. We
propose to use Alexa Fluor dyes as nano transmitters and receivers. These molecules, having
degrees of labelling up to 10, can attach in large numbers to proteins working as nano-
machines and thus create reliable MIMO-FRET communication channels. We present
experimental results of FRET efficiencies for nano-networks of Immunoglobulin G proteins
labelled with Alexa Fluor 405, 488 and 546 dyes. We calculate bit error rates for the FRET
communication channels under study. Finally, we extend the calculations for a general case of
MIMO (n,m) FRET channels. We show that FRET communication can be reliable even on
ranges larger than the Förster distances if multiple fluorophores are involved at both sides of
the channel.
REFERENCES
[Akyildiz-2008] I.F. Akyildiz, F. Brunetti, C. Blazquez, "NanoNetworking: A New
Communication Paradigm", Computer Networks (Elsevier) Journal, vol. 52, pp. 2260-2279,
August, 2008.
[Suda-2005] T. Suda, M. Moore, T. Nakano, R. Egashira, A. Enomoto, "Exploratory research
on molecular communication between nanomachines", Proceedings of Genetic and
Evolutionary Computation Conference (GECCO’05), June 2005.
[Nakano-2005] T. Nakano, T. Suda, M. Moore, R. Egashira, A. Enomoto, K. Arima,
Molecular communication for nanomachines using intercellular calcium signaling,
Proceedings of the Fifth IEEE Conference on Nanotechnology, June 2005, pp. 478–481.
[Gregori-2010] M. Gregori, I.F. Akyildiz, "A New NanoNetwork Architecture using
Flagellated Bacteria and Catalytic Nanomotors," IEEE Journal of Selected Areas in
Communications, vol. 28, pp. 612-619, May 2010.
[Moore-2006] M. Moore, A. Enomoto, T. Nakano, R. Egashira, T. Suda, A. Kayasuga,
H. Kojima, H. Sakakibara, K. Oiwa, "A design of a molecular communication system for
nanomachines using molecular motors", Proceedings of the Fourth Annual IEEE
International Conference on Pervasive Computing and Communications (PerCom’06), March
2006.
[Serreli-2007] V. Serreli, C. Lee, E.R. Kay, D.A. Leigh, A molecular information ratchet,
Nature 445: 523–527, 2007.
[Parcerisa-2009] L. Parcerisa, I.F. Akyildiz, "Molecular Communication Options for Long
Range Nanonetworks", Computer Networks (Elsevier) Journal, vol. 53, pp. 2753-2766, 2009.
[Nakano-2013] Tadashi Nakano, Andrew W. Eckford, Tokuko Haraguchi "Molecular
Communications", Cambridge University Press, 2013.
[Forster-1948] T. Förster "Zwischenmolekulare Energiewanderung und Fluoreszenz",
Annalen der Physik, vol. 437, no. 1–2, pp. 55–75, 1948.
[Kuscu-2012] M. Kuscu, O.B. Akan "A Physical Channel Model and Analysis for Nanoscale
Communications with Förster Resonance Energy Transfer (FRET),"IEEE Transactions on
Nanotechnology, vol. 11, no. 1, pp. 200-207, January 2012.
[Kuscu-2013] M. Kuscu, O.B. Akan "Multi-Step FRET-Based Long-Range Nanoscale
Communication Channel", IEEE Journal on Selected Areas in Communications (JSAC), vol.
31, no. 12, pp. 715-725, December 2013.
[Lakowicz-2006] Joseph R. Lakowicz "Principles of Fluorescence Spectroscopy 3 ed.",
Springer, 2006.
[Stryer-1978] L. Stryer "Fluorescence energy transfer as a spectroscopic ruler", Annual
Review of Biochemistry, vol. 47, pp. 819–846, 1978.
[Hannestad-2008] Jonas K. Hannestad, Peter Sandin, Bo Albinsson "Self-Assembled DNA
Photonic Wire for Long-Range Energy Transfer", Journal of the American Chemical Society
130(47):15889-95, 2008.
[Kasha-1950] M. Kasha "Characterization of electronic transitions in complex molecules",
Discussions of the Faraday Society 9:14–19, 1950.
[Tan-2008] Yih Horng Tan, Maozi Liu, Birte Nolting, Joan G. Go, Jacquelyn Gervay-Hague,
Gang-yu Liu, "A Nanoengineering Approach for Investigation and Regulation of Protein
Immobilization", ACS Nano 2 (11), pp. 2374–2384, 2008.
[Lifetechnologies-1] http://www.lifetechnologies.com/pl/en/home/references/molecular-
probes-the-handbook/tables/r0-values-for-some-alexa-fluor-dyes.html
[Lifetechnologies-2] http://www.lifetechnologies.com/europe-other/en/home/life-science/cell-
analysis/labeling-chemistry/fluorescence-spectraviewer.html